Social learning about predators: a review and prospectus

A. S. GRIFFIN ) 0 1 0 The author thanks Louis Lefebvre, Jeff Galef, and Douglas Chivers for comments on this paper. A.S.G. is supported by the Swiss National article should be addressed to A. S. Griffin, Department of Biology, McGill University , 1205 Dr Penfield Ave., Montreal , PQ, H3A 1B1 Canada ( 1 McGill University , Montreal , Quebec, Canada In comparison with social learning about food, social learning about predators has received little attention. Yet such research is of potential interest to students of animal cognition and conservation biologists. I summarize evidence for social learning about predators by fish, birds, eutherian mammals, and marsupials. I consider the proposal that this phenomenon is a case of S-S classical conditioning and suggest that evolution may have modified some of the properties of learning to accommodate for the requirements of learning socially about danger. I discuss some between-species differences in the properties of socially acquired predator avoidance and suggest that learning may be faster and more robust in species in which alarm behavior reliably predicts high predatory threat. Finally, I highlight how studies of socially acquired predator avoidance can inform the design of prerelease antipredator training programs for endangered species. - Intuitively, it seems that antipredator behavior should be fully functional upon a first encounter with danger. Indeed, some stimulus configurationsfor example, two black circlesare inherently aversive and trigger avoidance responses in animals with no prior experience of predators (Coss, 1978; Csnyi, 1985). On the other hand, there are reasons to predict that under some environmental conditions, antipredator behavior should be ontogenetically flexible. First, predation risk can vary in space and time. Learning allows quantitative levels of antipredator responses to be fine-tuned to local conditions (Lima & Dill, 1990). Second, environmental change can expose animals to previously unfamiliar predators, and learning allows novel dangers to be recognized (Berger, Swenson, & Persson, 2001). Third, community structures can change across generations. Under these conditions, recognition of stimuli, such as the alarm behavior of heterospecifics, may not evolve. Learning allows novel cues to become associated with predators. In keeping with these predictions, there is now abundant evidence that learning plays an important role both in the acquisition of antipredator responses and in the adjustment of preexisting ones. Most known examples of predator avoidance learning involve the use of social information. Unfortunately, the effects of direct experience with predators have received little attention. It is, therefore, difficult to tell whether the apparent importance of social influences on predator avoidance learning reflects an evolutionary trend favoring acquisition of risky information from others, rather than at ones own peril, or an explosion in the amount of research on social learning in the last 2 decades. Two patterns of social influence on predator avoidance have emerged. First, exposure to the alarm behavior of predatorexperienced social companions can enhance the frequency (Palleroni, 1999) or the specificity (Cheney & Seyfarth, 1990) of antipredator responses of juveniles or can cause response specificity to develop more quickly (Mateo, 1996; Mateo & Holmes, 1997). The second pattern of learning involves the acquisition of responses to previously unfamiliar stimuli and occurs in both juveniles and adults. This process has been termed observational conditioning (Cook, Mineka, Wolkenstein, & Laitsch, 1985), or releaser-induced recognition learning (Suboski, 1990), and is the focus of the present review. Socially acquired predator avoidance is a taxonomically widespread phenomenon that has been found in fish, birds, eutherians, and marsupials. The pattern of acquisition is similar across groups. Before learning, subjects show little or no response to a given stimulus. After that stimulus has been presented together with an alarm signal, however, it evokes an avoidance response. Several authors have noted the similarity between the process of predator avoidance acquisition and Pavlovian SS conditioning (Heyes, 1994; Mineka & Cook, 1993; Shettleworth, 1998; Suboski, 1990). Within this framework, the predatory cue is considered a conditional stimulus (CS) to which observers acquire avoidance responses after the stimulus has been presented in contiguity with an alarmed demonstrator, the unconditioned stimulus (US). Such an analysis is supported by the positive correlations between levels of demonstrator and observer alarm behavior during training and by the positive correlations between observer fear levels during and after training (Mineka & Cook, 1993). Socially acquired predator avoidance has recently been demonstrated in tammar wallabies (Macropus eugenii), an Australian macropodid marsupial (Griff in & Evans, 2003). This finding extended the existence of such learning to a new taxonomic group and provided the impetus for the present review. My objectives here are both to provide an overview of past work and to suggest new research approaches to mechanisms of socially transmitted predator avoidance. First, I will summarize the evidence for socially transmitted predator avoidance in fish, birds, and eutherian and marsupial mammals. I chose this taxonomic focus because the vast majority of studies on socially acquired predator avoidance have been conducted in these groups, even though such learning might occur in other taxa, such as amphibians and reptiles (Suboski, 1992). Second, I will examine whether the properties of learning support the idea that socially acquired predator avoidance is mediated by asocial learning mechanisms, rather than by some independent social learning process. Third, I will highlight some species differences and discuss possible reasons for them. Finally, I will briefly illustrate how the findings from basic studies of socially acquired predator avoidance can inform the design of prerelease antipredator training programs for endangered species. Predator avoidance learning in fish has been the focus of much basic research. There is some evidence that direct experience with predators (being startled or chased) can inculcate antipredator responses or enhance preexisting ones (Jrvi & Uglem, 1993). However, social cues seem to be particularly effective for triggering predator avoidance learning in this group. Social Stimuli That Trigger Learning in Observers The most intensively studied associative paradigm in f ish has involved paired presentations of unfamiliar predator cues with alarm pheromones. Von Frisch (1938) discovered that the skin of an injured fish releases chemical substances that evoke alarm responses in receivers. Shortly thereafter, Gz (1941) showed that these substances facilitate predator avoidance learning. Gz found that blinded Europea (...truncated)